solidjs / solid / 36098399060

74%

main: 89%

push

github

feat!: consolidate diagnostic codes; sourceNames follows dev in both compilers (#3646)

* feat!: consolidate diagnostic codes; sourceNames follows dev in both compilers

Diagnostic codes:
- HUGE_FAN_OUT absorbs WIDE_WRITE; the write-side attribution now ships as
  `data.write` on the same code, and the engine emits it directly.
- SERVER_ERROR_SANITIZED replaces SERVER_FN_ERROR_SANITIZED and
  SSR_ERROR_SANITIZED, distinguished by `data.source`.
- SSR_BOUNDARY_WATERFALL split out of ASYNC_WATERFALL (which drops `data.side`).
- AttributionOptions.wideWrites renamed to fanOut (`number | false`, default 250);
  new FanOutWrite type.

Compilers:
- `sourceNames` defaults to `dev` in both @solidjs/babel-plugin and
  @solidjs/compiler; object-form kinds default to `dev`; dev SSR keeps
  `createComponent` naming.
- Primitive naming stays with the native standalone pass; neither JSX compiler
  exposes a `primitives` key.

Co-authored-by: Claude via Cursor <noreply@cursor.com>
Co-authored-by: Cursor <cursoragent@cursor.com>

* test(bench): pin sourceNames off in the SSR bench lane

Vitest compiles with `dev: true`, and `sourceNames` now follows `dev` in
both compilers. For SSR output that keeps `createComponent(Comp, props,
"Name")` in place of the inlined `Comp(props)`, so the from-source (dev)
server runtime runs every component under a labelled transparent owner —
`observedComponent`: pool pop + field reset, `<Name>` string, closure,
try/finally per component instance. CodSpeed priced that at +24% on
`search-results: 50 items (renderToString)` and +5.9% on
`polymorphic-chain: chain-static` (~1000 component owners per render);
the remaining two deltas in the report touch benches with a single
component per render and are runtime-environment noise.

That owner is the diagnostics tier's cost, not the SSR runtime cost this
lane tracks. `vite.config.server-bench.mjs` now sets `sourceNames: false`
explicitly so the lane keeps measuring the production component shap... (continued)

605 of 894 branches covered (67.67%)

Branch coverage included in aggregate %.

883 of 1142 relevant lines covered (77.32%)

25.18 hits per line